Modeling Kidney Allocation: A Data-Driven Optimization Approach

In the United States, more than 90,000 candidates are currently waiting for kidney transplantation, with an annual increase of about 20,000 candidates. The current allocation policy poorly matches donors with recipients. We present a two-phase allocation policy that combines an integer programming-based learning phase and a datamining, real-time phase. Our policy outperforms the current system in multiple respects, such as increased life-year gained from kidney allocation and lower better match between organs and recipients.